Professional Documents
Culture Documents
Therapy
and Triple PPositive Parenting Program: A
Review and
Meta-Analysis
Rae Thomas Melanie J. Zimmer-Gembeck (2007)
Abstract We conducted a review and meta-analyses of 24 studies to evaluate and
compare the outcomes of two widely disseminated parenting interventionsParent-Child
Interaction
Therapy and Triple P-Positive Parenting Program. Participants
in all studies were caregivers and 3- to 12-year-old
children. In general, our analyses revealed positive effects
of both interventions, but effects varied depending on intervention
length, components, and source of outcome data.
Both interventions reduced parent-reported child behavior
and parenting problems. The effect sizes for PCIT were
large when outcomes of child and parent behaviors were
assessed with parent-report, with the exclusion of Abbreviated
PCIT, which had moderate effect sizes. All forms of
Triple P had moderate to large effects when outcomes were
parent-reported child behaviors and parenting, with the exception
of Media Triple P, which had small effects. PCIT
and an enhanced version of Triple P were associated with
improvements in observed child behaviors. These findings
provide information about the relative efficacy of two programs
that have received substantial funding in the USA and
Australia, and findings should assist in making decisions
about allocations of funding and dissemination of these parenting
interventions in the future.
Behavioral parent training interventions are widely used because
the evidence shows they are efficacious (Serketich &
Dumas, 1996). Although parenting interventions focusing on
child behavior management are widespread there are a variety
of theoretical foundations and delivery formats for these
interventions (Kumpfer & Alvarado, 2003). Nevertheless,
behavioral parent training, where parents participate in skills
training, is included in many interventions with demonstrated
efficacy (Serketich & Dumas, 1996). Often, behavioral parent
training involves discussing differential reinforcement
and timeout procedures, and an emphasis is placed on the
role the parent plays in the development and maintenance of
child problem behaviors (Hollenstein, Granic, Stoolmiller,
& Snyder, 2004). Yet, interventions might only include parents,
might involve family skills training, where the parents
and children are taught skills and have time to practice these
together with a therapist, or interventions may involve family
therapy in which the family unit receives intervention with
less of the content directed at skill development.
Clinical benefits of behavioral parent training interventions
were described in one meta-analysis of 26 published
studies (Sereketich & Dumas, 1996). In another systematic
review (Barlow & Stewart-Brown, 2000), 16 randomized
Mt-post and Mc-post were the mean values for the treatment
group (t-post) or the comparison group (c-post) at posttest
or at follow-up,
Mt-pre and Mc-pre were the mean values for the treatment
group (t-pre) or the comparison group (c-pre) at pre-test,
SDc-pre and SDc-pre were the standard deviations at pretest.
Hence, dig was the treatment effect size net of the comparison
group effect size. When dig was positive this indicated
a larger effect in the treatment group than in the comparison
group. When dig was negative this indicated a larger effect
in the comparison group than in the treatment group.
We used pre-test SDs for all calculations, as suggested
by Becker (1988), because pre-test SDs are often more
consistent across studies than post-test SDs. The use of pretest
SDs to estimate effect sizes, therefore, results in estimates
that are more comparable across different experimental
manipulations (in this case, between studies of PCIT and
Triple P). Post-test SDs have been found to be less consistent
than pre-test SDs as a result of different experimental
manipulations. We kept single group effect sizes separate
from independent group effect sizes, because there is continuing
debate about when these can be combined, and because
single group effect sizes are often larger than independent
group effect sizes (see Carlson & Schmidt, 1999; Morris &
DeShon, 2002). Although there are many reasons that single
group and independent group effect sizes might differ and
both study designs can have their own biases, one reason
single group effect sizes often are larger is because they do
not take into account the time effect (e.g., history or maturation
effect) that can be partly accounted for by a comparison
group (Morris & DeShon, 2002).
PCIT effect sizes
Thirteen studies were included in calculations of SGPP effect
sizes; eight studies were included in calculations of IGPP
effect sizes. Four studies were available for calculations of
SGPF and IGPF effects.We calculated the following for child
behavioral outcomes: 65 SGPP effects, 30 SGPF (4 month to
1 year after treatment) effects, 49 IGPP effect sizes, and 14
IGPF effect sizes. For parenting behaviors, we calculated 47
SGPP effects, 14 SGPF (4 month to 1 year after treatment)
effects, 54 IGPP effects, and 16 IGPF effects.
Triple P effect sizes
Eleven studies were included in the calculations of SGPP;
whereas seven studies were included in calculations of IGPP
There were no studies with comparison groups that included
a follow-up assessment.We calculated the following for child
behavioral outcomes: 78 SGPP effects, 27 SGPF (4 month to
1 year after treatment) effects, and 53 IGPP effects. For parenting
behaviors, we calculated 77 SGPP effects, 55 SGPF (4
month to 1 year after treatment) effects, and 50 IGPP effects.
Reduction of effect sizes and final analysis
Effect sizes were categorized by purpose of the measure (to
assess child or parent), method (questionnaire versus observation)
and reporter (mother, father, teacher). In some cases,
effect sizes in the same category within a study were averaged
after accounting for repetition from multiple studies
from the same sample. This averaging was done to reduce
bias that might be introduced from studies that used more
measures, and findings reported in more than one publication
Media
There was no significant effect of Media Triple P on parenting
based on self-report questionnaire.
Results: Comparison of PCIT and Triple P
We used categorical modeling to compare the effect sizes
for PCIT and Triple P (Johnson, 1989). We conducted analyses
to compare SGPP effect sizes, and compared IGPP for
Standard PCIT (compared to waitlist) to the multiple forms
of Triple P (compared to waitlist).
Single group pre-treatment/post-treatment (SGPP)
effect sizes
For SGPP, PCIT had larger effects of parent report of child
negative behaviors, but not when effect sizes were based
on observed child behaviors. More specifically, there was
a larger improvement for study families in PCIT than in
Triple P for parent-report of child behavior problems, 1.31
vs. .73, p<.001 (see Tables 25). However, there was no
difference in SGPP effect size for observed child negative
behavior, .54 vs. .31, p = .27.
Findings when comparing SGPP for parenting problems
showed that PCIT had larger effects for both parent report
and observed parent negative behaviors. There was a larger
improvement for study families in PCIT than in Triple P
when outcomes were based on parent report of parenting
problems, 1.11 vs. .70, p<.001. There also was a larger
improvement in observed parent negative behavior for PCIT
than Triple P, 1.46 vs. .19, p<.001.
Independent groups pre-treatment/post-treatment
(IGPP) effect sizes
Child negative behaviors
Figure 1 shows IGPP effect sizes and confidence intervals
for the four forms of PCIT and five forms of Triple P when
compared to waitlist. The effects in Fig. 1 are based on
parent-reports of child negative behavior. As shown, there
were large effect sizes for child negative behavior in all forms
of PCIT, except the Abbreviated version, which did not have
a significant effect on child negative behavior. There were
medium to large effects for all forms of Triple P.
As can be seen by comparing the confidence intervals in
Fig. 1, results of our analyses that compared Standard PCIT
to multiple forms of Triple P showed differences in PCIT and
some forms of Triple P for parent report of child negative
behaviors. The effect size for PCIT, 1.45,was significantly
larger than the effect sizes for Self-directed, .51, p<.001,
Group, .67, p<.01, and Individual Triple P, .69, p<.01,
but the PCIT effect size was not larger than the effect size
for Enhanced, .96, or Media Triple P, .79. However, in
contrast to these findings, for IGPP based on observed child
negative behaviors, there were no differences in the effect
size for Standard PCIT, .11, compared to Triple P in the
Self-directed, .02, Individual, .22, and Enhanced, .46,
forms (figure not provided).
Parent negative behaviors
Figure 2 summarizes IGPP effect sizes based on parent reports
of parent negative behavior, and illustrates that all forms
of PCIT and Triple P, except Media Triple P, had medium or
large effects. Media Triple P did not have a significant effect
on self-reported negative parent behavior. All other interventions
reduced negative parenting of those in treatment